MKXJun / SMPL-Shape-Control

Semantically Controllable SMPL Shape

Accurate height control without variation of arm span, inseam height, chest, waist and hip

Dependencies for demo

pytorch
numpy
trimesh
pyrender
smplx

You need to put SMPLX_MALE.npz into data/body_models/smplx folder.

Demo

• 1: modify height
• 2: modify inseam height
• 3: modify arm span
• 4: modify chest
• 5: modify waist
• 6: modify hip
• -: decrease 0.5cm
• +: increase 0.5cm
• space: new random shape

Roadmap

• SMPL-X shape control
• SMPL shape control
• SMPL-H shape control
• Female shape control

when paper is available(maybe a long long time):

• SMPL shape generation
• semantically controllable SMPL shape parameter generation

Usage

import torch
import numpy as np

# ...

height = torch.from_numpy(np.load('data/smplx_params/smplx_params_height_1cm_b10.npz')['betas']) # [1, 10]
modified_betas = betas + height * 3.0  # increase height by 3.0 cm